| { | |
| "best_metric": 0.8433638471389473, | |
| "best_model_checkpoint": "result/my-unsup-simcse-roberta-base-d0.1-rs192-std0.1-t0.05", | |
| "epoch": 1.0, | |
| "global_step": 15626, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "eval_avg_sts": 0.7572535655220898, | |
| "eval_sickr_spearman": 0.7054352844388353, | |
| "eval_stsb_spearman": 0.8090718466053444, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "eval_avg_sts": 0.7614025514877952, | |
| "eval_sickr_spearman": 0.7051708251943171, | |
| "eval_stsb_spearman": 0.8176342777812734, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "eval_avg_sts": 0.7636217659471919, | |
| "eval_sickr_spearman": 0.7079373205749496, | |
| "eval_stsb_spearman": 0.8193062113194342, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.680020478689364e-06, | |
| "loss": 0.2125, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "eval_avg_sts": 0.7657425201047614, | |
| "eval_sickr_spearman": 0.7097071706007412, | |
| "eval_stsb_spearman": 0.8217778696087817, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "eval_avg_sts": 0.7674773834562932, | |
| "eval_sickr_spearman": 0.7100151940542637, | |
| "eval_stsb_spearman": 0.8249395728583228, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "eval_avg_sts": 0.7670247636071179, | |
| "eval_sickr_spearman": 0.7053980603352208, | |
| "eval_stsb_spearman": 0.828651466879015, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "eval_avg_sts": 0.7642544693075077, | |
| "eval_sickr_spearman": 0.7042453619318076, | |
| "eval_stsb_spearman": 0.8242635766832077, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.36004095737873e-06, | |
| "loss": 0.0005, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "eval_avg_sts": 0.7706729186771359, | |
| "eval_sickr_spearman": 0.7109765365495476, | |
| "eval_stsb_spearman": 0.8303693008047242, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "eval_avg_sts": 0.7723601548256038, | |
| "eval_sickr_spearman": 0.7154485202800445, | |
| "eval_stsb_spearman": 0.8292717893711632, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "eval_avg_sts": 0.7715988622174201, | |
| "eval_sickr_spearman": 0.7117784157880568, | |
| "eval_stsb_spearman": 0.8314193086467835, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "eval_avg_sts": 0.7699957963614845, | |
| "eval_sickr_spearman": 0.7098231176796127, | |
| "eval_stsb_spearman": 0.8301684750433563, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.040061436068092e-06, | |
| "loss": 0.0003, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "eval_avg_sts": 0.7732021256956165, | |
| "eval_sickr_spearman": 0.7176254338905241, | |
| "eval_stsb_spearman": 0.8287788175007089, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "eval_avg_sts": 0.7706739860482326, | |
| "eval_sickr_spearman": 0.7126164144148476, | |
| "eval_stsb_spearman": 0.8287315576816174, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "eval_avg_sts": 0.7719707314787304, | |
| "eval_sickr_spearman": 0.7146769966976441, | |
| "eval_stsb_spearman": 0.8292644662598168, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "eval_avg_sts": 0.7692985063340654, | |
| "eval_sickr_spearman": 0.7109207724407779, | |
| "eval_stsb_spearman": 0.827676240227353, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 8.720081914757458e-06, | |
| "loss": 0.0002, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "eval_avg_sts": 0.771889783054714, | |
| "eval_sickr_spearman": 0.7135132511409, | |
| "eval_stsb_spearman": 0.8302663149685279, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "eval_avg_sts": 0.7615509683620076, | |
| "eval_sickr_spearman": 0.6917990552038427, | |
| "eval_stsb_spearman": 0.8313028815201725, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "eval_avg_sts": 0.7642814292065848, | |
| "eval_sickr_spearman": 0.6928178503087533, | |
| "eval_stsb_spearman": 0.8357450081044163, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "eval_avg_sts": 0.7664718523224301, | |
| "eval_sickr_spearman": 0.6978614041463637, | |
| "eval_stsb_spearman": 0.8350823004984966, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 8.400102393446819e-06, | |
| "loss": 0.0002, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "eval_avg_sts": 0.7669677380011899, | |
| "eval_sickr_spearman": 0.6978531908280179, | |
| "eval_stsb_spearman": 0.8360822851743619, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "eval_avg_sts": 0.767509277729593, | |
| "eval_sickr_spearman": 0.697946947538025, | |
| "eval_stsb_spearman": 0.837071607921161, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "eval_avg_sts": 0.7668651085967232, | |
| "eval_sickr_spearman": 0.6985169326187911, | |
| "eval_stsb_spearman": 0.8352132845746553, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "eval_avg_sts": 0.7668897285698411, | |
| "eval_sickr_spearman": 0.6986797580526662, | |
| "eval_stsb_spearman": 0.835099699087016, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 8.080122872136184e-06, | |
| "loss": 0.0002, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "eval_avg_sts": 0.7678085026409753, | |
| "eval_sickr_spearman": 0.6984642425005135, | |
| "eval_stsb_spearman": 0.8371527627814372, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "eval_avg_sts": 0.7692156947271456, | |
| "eval_sickr_spearman": 0.7026153304423024, | |
| "eval_stsb_spearman": 0.8358160590119886, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "eval_avg_sts": 0.768843154466003, | |
| "eval_sickr_spearman": 0.705912809649333, | |
| "eval_stsb_spearman": 0.8317734992826732, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "eval_avg_sts": 0.7672560021396605, | |
| "eval_sickr_spearman": 0.7075799211491486, | |
| "eval_stsb_spearman": 0.8269320831301724, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 7.760143350825547e-06, | |
| "loss": 0.0002, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "eval_avg_sts": 0.7665000851835462, | |
| "eval_sickr_spearman": 0.7056194356817491, | |
| "eval_stsb_spearman": 0.8273807346853432, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "eval_avg_sts": 0.7675835380050031, | |
| "eval_sickr_spearman": 0.706044462898375, | |
| "eval_stsb_spearman": 0.8291226131116314, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "eval_avg_sts": 0.766064422853222, | |
| "eval_sickr_spearman": 0.7050694795702827, | |
| "eval_stsb_spearman": 0.8270593661361613, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "eval_avg_sts": 0.7680428355810465, | |
| "eval_sickr_spearman": 0.7066562350373923, | |
| "eval_stsb_spearman": 0.8294294361247007, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 7.440163829514912e-06, | |
| "loss": 0.0003, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "eval_avg_sts": 0.7706781402051571, | |
| "eval_sickr_spearman": 0.7060189583835114, | |
| "eval_stsb_spearman": 0.8353373220268029, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "eval_avg_sts": 0.7704250512723954, | |
| "eval_sickr_spearman": 0.7065240054151332, | |
| "eval_stsb_spearman": 0.8343260971296576, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "eval_avg_sts": 0.7676150118719631, | |
| "eval_sickr_spearman": 0.7018462564460752, | |
| "eval_stsb_spearman": 0.833383767297851, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "eval_avg_sts": 0.7672794331395227, | |
| "eval_sickr_spearman": 0.7023547617170005, | |
| "eval_stsb_spearman": 0.832204104562045, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 7.120184308204276e-06, | |
| "loss": 0.0003, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "eval_avg_sts": 0.7603546899710306, | |
| "eval_sickr_spearman": 0.6936834187877694, | |
| "eval_stsb_spearman": 0.8270259611542918, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "eval_avg_sts": 0.7604700472645542, | |
| "eval_sickr_spearman": 0.6935327932536596, | |
| "eval_stsb_spearman": 0.8274073012754489, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "eval_avg_sts": 0.761349850549489, | |
| "eval_sickr_spearman": 0.6934180469523239, | |
| "eval_stsb_spearman": 0.8292816541466541, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "eval_avg_sts": 0.7624410643103536, | |
| "eval_sickr_spearman": 0.6947914482668454, | |
| "eval_stsb_spearman": 0.830090680353862, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 6.800204786893639e-06, | |
| "loss": 0.0001, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "eval_avg_sts": 0.7625184502556002, | |
| "eval_sickr_spearman": 0.6949592209041686, | |
| "eval_stsb_spearman": 0.8300776796070318, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "eval_avg_sts": 0.7620233259863773, | |
| "eval_sickr_spearman": 0.6943664690813208, | |
| "eval_stsb_spearman": 0.829680182891434, | |
| "step": 5125 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "eval_avg_sts": 0.7688935367788103, | |
| "eval_sickr_spearman": 0.7047914275240574, | |
| "eval_stsb_spearman": 0.8329956460335631, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "eval_avg_sts": 0.7721471528975761, | |
| "eval_sickr_spearman": 0.7084637894778126, | |
| "eval_stsb_spearman": 0.8358305163173394, | |
| "step": 5375 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 6.480225265583003e-06, | |
| "loss": 0.0002, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "eval_avg_sts": 0.7736568332643656, | |
| "eval_sickr_spearman": 0.7087314187750259, | |
| "eval_stsb_spearman": 0.8385822477537053, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "eval_avg_sts": 0.7734070993186029, | |
| "eval_sickr_spearman": 0.7073194965171511, | |
| "eval_stsb_spearman": 0.8394947021200546, | |
| "step": 5625 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "eval_avg_sts": 0.7727328180179123, | |
| "eval_sickr_spearman": 0.7090136014959747, | |
| "eval_stsb_spearman": 0.83645203453985, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "eval_avg_sts": 0.7702894469668818, | |
| "eval_sickr_spearman": 0.7065904804595234, | |
| "eval_stsb_spearman": 0.8339884134742402, | |
| "step": 5875 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.1602457442723675e-06, | |
| "loss": 0.0001, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "eval_avg_sts": 0.7673639531247151, | |
| "eval_sickr_spearman": 0.7063840427855425, | |
| "eval_stsb_spearman": 0.8283438634638877, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "eval_avg_sts": 0.7656242558893468, | |
| "eval_sickr_spearman": 0.7046311957696599, | |
| "eval_stsb_spearman": 0.8266173160090337, | |
| "step": 6125 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "eval_avg_sts": 0.7678162328535889, | |
| "eval_sickr_spearman": 0.7065671853753259, | |
| "eval_stsb_spearman": 0.8290652803318519, | |
| "step": 6250 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "eval_avg_sts": 0.7659380803169125, | |
| "eval_sickr_spearman": 0.7034162970898843, | |
| "eval_stsb_spearman": 0.8284598635439405, | |
| "step": 6375 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 5.840266222961732e-06, | |
| "loss": 0.0002, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "eval_avg_sts": 0.7674929110456945, | |
| "eval_sickr_spearman": 0.7043084267679958, | |
| "eval_stsb_spearman": 0.8306773953233932, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "eval_avg_sts": 0.7674747228197092, | |
| "eval_sickr_spearman": 0.7030638928986329, | |
| "eval_stsb_spearman": 0.8318855527407855, | |
| "step": 6625 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "eval_avg_sts": 0.7669269125478364, | |
| "eval_sickr_spearman": 0.7022142707452941, | |
| "eval_stsb_spearman": 0.8316395543503788, | |
| "step": 6750 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "eval_avg_sts": 0.7663176164301795, | |
| "eval_sickr_spearman": 0.7013282410170653, | |
| "eval_stsb_spearman": 0.8313069918432938, | |
| "step": 6875 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.520286701651095e-06, | |
| "loss": 0.0001, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "eval_avg_sts": 0.7679613274313469, | |
| "eval_sickr_spearman": 0.705412661790058, | |
| "eval_stsb_spearman": 0.8305099930726358, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "eval_avg_sts": 0.7677041476146869, | |
| "eval_sickr_spearman": 0.7044645758787711, | |
| "eval_stsb_spearman": 0.8309437193506026, | |
| "step": 7125 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "eval_avg_sts": 0.771967747158485, | |
| "eval_sickr_spearman": 0.7112394107677183, | |
| "eval_stsb_spearman": 0.8326960835492517, | |
| "step": 7250 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "eval_avg_sts": 0.7727980509536904, | |
| "eval_sickr_spearman": 0.7122028666314655, | |
| "eval_stsb_spearman": 0.8333932352759152, | |
| "step": 7375 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.200307180340458e-06, | |
| "loss": 0.0002, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "eval_avg_sts": 0.7716808664260861, | |
| "eval_sickr_spearman": 0.7073300153283661, | |
| "eval_stsb_spearman": 0.8360317175238062, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "eval_avg_sts": 0.7749475591557311, | |
| "eval_sickr_spearman": 0.7102010744168291, | |
| "eval_stsb_spearman": 0.8396940438946332, | |
| "step": 7625 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "eval_avg_sts": 0.7741400965390657, | |
| "eval_sickr_spearman": 0.7103573195898074, | |
| "eval_stsb_spearman": 0.8379228734883241, | |
| "step": 7750 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "eval_avg_sts": 0.7742400629462907, | |
| "eval_sickr_spearman": 0.7103860902195688, | |
| "eval_stsb_spearman": 0.8380940356730125, | |
| "step": 7875 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.8803276590298225e-06, | |
| "loss": 0.0001, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "eval_avg_sts": 0.7737984133316254, | |
| "eval_sickr_spearman": 0.7100542913708344, | |
| "eval_stsb_spearman": 0.8375425352924163, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "eval_avg_sts": 0.7736256294772575, | |
| "eval_sickr_spearman": 0.7117209225596355, | |
| "eval_stsb_spearman": 0.8355303363948796, | |
| "step": 8125 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "eval_avg_sts": 0.7739892471083305, | |
| "eval_sickr_spearman": 0.7121400419507844, | |
| "eval_stsb_spearman": 0.8358384522658765, | |
| "step": 8250 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "eval_avg_sts": 0.771587372188157, | |
| "eval_sickr_spearman": 0.7086112929903291, | |
| "eval_stsb_spearman": 0.834563451385985, | |
| "step": 8375 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.560348137719187e-06, | |
| "loss": 0.0001, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "eval_avg_sts": 0.7717136828796769, | |
| "eval_sickr_spearman": 0.7081623943162886, | |
| "eval_stsb_spearman": 0.8352649714430652, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "eval_avg_sts": 0.7700399485422347, | |
| "eval_sickr_spearman": 0.7062681917688737, | |
| "eval_stsb_spearman": 0.8338117053155956, | |
| "step": 8625 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "eval_avg_sts": 0.7655558973491845, | |
| "eval_sickr_spearman": 0.7032426646581855, | |
| "eval_stsb_spearman": 0.8278691300401836, | |
| "step": 8750 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "eval_avg_sts": 0.7691329874077902, | |
| "eval_sickr_spearman": 0.7023757513083291, | |
| "eval_stsb_spearman": 0.8358902235072514, | |
| "step": 8875 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 4.24036861640855e-06, | |
| "loss": 0.0001, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "eval_avg_sts": 0.7696286255223032, | |
| "eval_sickr_spearman": 0.7031883895135604, | |
| "eval_stsb_spearman": 0.8360688615310459, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "eval_avg_sts": 0.7696384100854612, | |
| "eval_sickr_spearman": 0.7036895940570672, | |
| "eval_stsb_spearman": 0.835587226113855, | |
| "step": 9125 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "eval_avg_sts": 0.7688416464364851, | |
| "eval_sickr_spearman": 0.7023802182007628, | |
| "eval_stsb_spearman": 0.8353030746722075, | |
| "step": 9250 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "eval_avg_sts": 0.7660994859121804, | |
| "eval_sickr_spearman": 0.7051366750811947, | |
| "eval_stsb_spearman": 0.8270622967431659, | |
| "step": 9375 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.920389095097914e-06, | |
| "loss": 0.0002, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "eval_avg_sts": 0.7673638146997882, | |
| "eval_sickr_spearman": 0.7064614138295504, | |
| "eval_stsb_spearman": 0.828266215570026, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "eval_avg_sts": 0.7688152733727235, | |
| "eval_sickr_spearman": 0.707200523478889, | |
| "eval_stsb_spearman": 0.830430023266558, | |
| "step": 9625 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "eval_avg_sts": 0.7697361419751771, | |
| "eval_sickr_spearman": 0.7075413521746937, | |
| "eval_stsb_spearman": 0.8319309317756607, | |
| "step": 9750 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "eval_avg_sts": 0.7718885121529019, | |
| "eval_sickr_spearman": 0.7071343846522086, | |
| "eval_stsb_spearman": 0.8366426396535952, | |
| "step": 9875 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 3.600409573787278e-06, | |
| "loss": 0.0001, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "eval_avg_sts": 0.7727693021998165, | |
| "eval_sickr_spearman": 0.7095777267823654, | |
| "eval_stsb_spearman": 0.8359608776172676, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "eval_avg_sts": 0.7717316927837671, | |
| "eval_sickr_spearman": 0.7084112434528395, | |
| "eval_stsb_spearman": 0.8350521421146947, | |
| "step": 10125 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "eval_avg_sts": 0.7720632550487434, | |
| "eval_sickr_spearman": 0.7069583986965393, | |
| "eval_stsb_spearman": 0.8371681114009475, | |
| "step": 10250 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "eval_avg_sts": 0.7716685440161564, | |
| "eval_sickr_spearman": 0.7068186762224559, | |
| "eval_stsb_spearman": 0.836518411809857, | |
| "step": 10375 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 3.280430052476642e-06, | |
| "loss": 0.0001, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "eval_avg_sts": 0.7706994083764094, | |
| "eval_sickr_spearman": 0.7044442106917614, | |
| "eval_stsb_spearman": 0.8369546060610572, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "eval_avg_sts": 0.770376770388492, | |
| "eval_sickr_spearman": 0.704766259226904, | |
| "eval_stsb_spearman": 0.83598728155008, | |
| "step": 10625 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "eval_avg_sts": 0.7700945307869602, | |
| "eval_sickr_spearman": 0.7045319154829873, | |
| "eval_stsb_spearman": 0.8356571460909332, | |
| "step": 10750 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "eval_avg_sts": 0.7701867471043669, | |
| "eval_sickr_spearman": 0.7049941187721263, | |
| "eval_stsb_spearman": 0.8353793754366073, | |
| "step": 10875 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 2.960450531166006e-06, | |
| "loss": 0.0001, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "eval_avg_sts": 0.7700153990116653, | |
| "eval_sickr_spearman": 0.7048114084622558, | |
| "eval_stsb_spearman": 0.8352193895610748, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "eval_avg_sts": 0.7693725150897858, | |
| "eval_sickr_spearman": 0.7057958058862298, | |
| "eval_stsb_spearman": 0.8329492242933417, | |
| "step": 11125 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "eval_avg_sts": 0.7696360829638863, | |
| "eval_sickr_spearman": 0.7062946088746646, | |
| "eval_stsb_spearman": 0.832977557053108, | |
| "step": 11250 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "eval_avg_sts": 0.7709027781052127, | |
| "eval_sickr_spearman": 0.7053959469667576, | |
| "eval_stsb_spearman": 0.8364096092436679, | |
| "step": 11375 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 2.640471009855369e-06, | |
| "loss": 0.0002, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "eval_avg_sts": 0.7705889705773711, | |
| "eval_sickr_spearman": 0.7048775472889359, | |
| "eval_stsb_spearman": 0.8363003938658062, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "eval_avg_sts": 0.7756785758354756, | |
| "eval_sickr_spearman": 0.7125353379156201, | |
| "eval_stsb_spearman": 0.8388218137553312, | |
| "step": 11625 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "eval_avg_sts": 0.775711044555684, | |
| "eval_sickr_spearman": 0.7125440315449804, | |
| "eval_stsb_spearman": 0.8388780575663876, | |
| "step": 11750 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "eval_avg_sts": 0.7756594965556575, | |
| "eval_sickr_spearman": 0.710795363234923, | |
| "eval_stsb_spearman": 0.840523629876392, | |
| "step": 11875 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.3204914885447333e-06, | |
| "loss": 0.0002, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "eval_avg_sts": 0.778271737124629, | |
| "eval_sickr_spearman": 0.7131796271103109, | |
| "eval_stsb_spearman": 0.8433638471389473, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "eval_avg_sts": 0.7778824941586701, | |
| "eval_sickr_spearman": 0.7129357251772082, | |
| "eval_stsb_spearman": 0.842829263140132, | |
| "step": 12125 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "eval_avg_sts": 0.7791778033988318, | |
| "eval_sickr_spearman": 0.7154212866455291, | |
| "eval_stsb_spearman": 0.8429343201521345, | |
| "step": 12250 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "eval_avg_sts": 0.7784491802919915, | |
| "eval_sickr_spearman": 0.7169989642344666, | |
| "eval_stsb_spearman": 0.8398993963495163, | |
| "step": 12375 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.000511967234097e-06, | |
| "loss": 0.0003, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "eval_avg_sts": 0.7782960644263726, | |
| "eval_sickr_spearman": 0.7171246616269303, | |
| "eval_stsb_spearman": 0.8394674672258147, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "eval_avg_sts": 0.7777553880232895, | |
| "eval_sickr_spearman": 0.7163744638535685, | |
| "eval_stsb_spearman": 0.8391363121930105, | |
| "step": 12625 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "eval_avg_sts": 0.7768079269313797, | |
| "eval_sickr_spearman": 0.7156439108006947, | |
| "eval_stsb_spearman": 0.8379719430620648, | |
| "step": 12750 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "eval_avg_sts": 0.7769940996454496, | |
| "eval_sickr_spearman": 0.7154824302376598, | |
| "eval_stsb_spearman": 0.8385057690532396, | |
| "step": 12875 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.6805324459234608e-06, | |
| "loss": 0.0002, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "eval_avg_sts": 0.7773811376862301, | |
| "eval_sickr_spearman": 0.7153101426768013, | |
| "eval_stsb_spearman": 0.839452132695659, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "eval_avg_sts": 0.7765844510293946, | |
| "eval_sickr_spearman": 0.713703022022682, | |
| "eval_stsb_spearman": 0.8394658800361071, | |
| "step": 13125 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "eval_avg_sts": 0.7754935189329493, | |
| "eval_sickr_spearman": 0.7128497975367355, | |
| "eval_stsb_spearman": 0.838137240329163, | |
| "step": 13250 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "eval_avg_sts": 0.7753794499527557, | |
| "eval_sickr_spearman": 0.7128604124101533, | |
| "eval_stsb_spearman": 0.837898487495358, | |
| "step": 13375 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.3605529246128248e-06, | |
| "loss": 0.0001, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "eval_avg_sts": 0.7753019292848522, | |
| "eval_sickr_spearman": 0.7128690099773107, | |
| "eval_stsb_spearman": 0.8377348485923936, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "eval_avg_sts": 0.7751785379635656, | |
| "eval_sickr_spearman": 0.7127431615818965, | |
| "eval_stsb_spearman": 0.8376139143452346, | |
| "step": 13625 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "eval_avg_sts": 0.7751092404866357, | |
| "eval_sickr_spearman": 0.712538940248228, | |
| "eval_stsb_spearman": 0.8376795407250434, | |
| "step": 13750 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "eval_avg_sts": 0.7748652873780293, | |
| "eval_sickr_spearman": 0.7121167948976883, | |
| "eval_stsb_spearman": 0.8376137798583702, | |
| "step": 13875 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0405734033021888e-06, | |
| "loss": 0.0, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "eval_avg_sts": 0.7749284965089291, | |
| "eval_sickr_spearman": 0.7121870643990923, | |
| "eval_stsb_spearman": 0.8376699286187659, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "eval_avg_sts": 0.7736810149784585, | |
| "eval_sickr_spearman": 0.7101105357906183, | |
| "eval_stsb_spearman": 0.8372514941662987, | |
| "step": 14125 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "eval_avg_sts": 0.7736952566877616, | |
| "eval_sickr_spearman": 0.7101496811382905, | |
| "eval_stsb_spearman": 0.8372408322372328, | |
| "step": 14250 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "eval_avg_sts": 0.7737241698204669, | |
| "eval_sickr_spearman": 0.7102500181091946, | |
| "eval_stsb_spearman": 0.837198321531739, | |
| "step": 14375 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 7.205938819915525e-07, | |
| "loss": 0.0001, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "eval_avg_sts": 0.7737795026917761, | |
| "eval_sickr_spearman": 0.7103373866827106, | |
| "eval_stsb_spearman": 0.8372216187008416, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "eval_avg_sts": 0.7736836444786954, | |
| "eval_sickr_spearman": 0.7101872414596151, | |
| "eval_stsb_spearman": 0.8371800474977756, | |
| "step": 14625 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "eval_avg_sts": 0.7734449184616194, | |
| "eval_sickr_spearman": 0.7102963681220825, | |
| "eval_stsb_spearman": 0.8365934688011564, | |
| "step": 14750 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "eval_avg_sts": 0.7733655720089518, | |
| "eval_sickr_spearman": 0.7101887784548611, | |
| "eval_stsb_spearman": 0.8365423655630426, | |
| "step": 14875 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 4.0061436068091647e-07, | |
| "loss": 0.0001, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "eval_avg_sts": 0.7735006567182534, | |
| "eval_sickr_spearman": 0.7106506935573914, | |
| "eval_stsb_spearman": 0.8363506198791154, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "eval_avg_sts": 0.7733751888661335, | |
| "eval_sickr_spearman": 0.710443343292483, | |
| "eval_stsb_spearman": 0.8363070344397839, | |
| "step": 15125 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "eval_avg_sts": 0.7736170866924498, | |
| "eval_sickr_spearman": 0.7105904625561881, | |
| "eval_stsb_spearman": 0.8366437108287115, | |
| "step": 15250 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "eval_avg_sts": 0.7738120865784781, | |
| "eval_sickr_spearman": 0.7107394070017475, | |
| "eval_stsb_spearman": 0.8368847661552088, | |
| "step": 15375 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 8.06348393702803e-08, | |
| "loss": 0.0002, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "eval_avg_sts": 0.7737627481564555, | |
| "eval_sickr_spearman": 0.710675861854545, | |
| "eval_stsb_spearman": 0.8368496344583661, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_avg_sts": 0.7737442273397684, | |
| "eval_sickr_spearman": 0.7106493006554496, | |
| "eval_stsb_spearman": 0.8368391540240873, | |
| "step": 15625 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "step": 15626, | |
| "train_runtime": 3862.446, | |
| "train_samples_per_second": 4.046 | |
| } | |
| ], | |
| "max_steps": 15626, | |
| "num_train_epochs": 1, | |
| "total_flos": 48090729552850944, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |